It was 2018, and the world as we knew it--or rather, how we knew it--teetered on a precipice. Against a rising drone of misinformation, The New York Times, the BBC, Good Morning America, and just about everyone else sounded the alarm over a new strain of fake but highly realistic videos. Using artificial intelligence, bad actors could manipulate someone's voice and face in recorded footage almost like a virtual puppet and pass the product off as real. In a famous example engineered by BuzzFeed, Barack Obama seemed to say, "President Trump is a total and complete dipshit." Synthetic photos, audio, and videos, collectively dubbed "deepfakes," threatened to destabilize society and push us into a full-blown "infocalypse."

This seems the only suitable word to assess the huge amount of data being generated due to the ensuing COVID 19 pandemic and the global lockdown caused by it. We can broadly classify the data into two categories – Deliberate and Non-Deliberate. The first category of the data is being generated by governments as part of their response plan to the pandemic while the second category of data is being automatically generated due to the global lockdown. As the governments have well-defined objectives to create and use the data they are generating to control the outbreak of COVID 19 in their respective territories, this category of data is immediately being used in their outbreak response plans such as communication campaigns, diseases prevention, social distancing, awareness campaigns, diagnosis, prognosis, and treatment with the help of AI (Artificial Intelligence) based technological innovations particularly mobile apps, dashboard, websites, etc. In addition to the urgent disease containment plans, the first category of data will also be crucial for assessing health systems, developing pandemic/epidemic/outbreak resilience plans and assessing economic impacts to improve future resilience. However, the collection and use of the second category of data is likely to guide the national and global policies for the years from transport planning, supply chain management, global warming, carbon emission, climate change, biodiversity, regional cooperation, geopolitics and much more.

Pretrained language models (PLMs) often fail to fairly represent target users from certain world regions because of the under-representation of those regions in training datasets. With recent PLMs trained on enormous data sources, quantifying their potential biases is difficult, due to their black-box nature and the sheer scale of the data sources. In this work, we devise an approach to study the geographic bias (and knowledge) present in PLMs, proposing a Geographic-Representation Probing Framework adopting a self-conditioning method coupled with entity-country mappings. Our findings suggest PLMs' representations map surprisingly well to the physical world in terms of country-to-country associations, but this knowledge is unequally shared across languages. Last, we explain how large PLMs despite exhibiting notions of geographical proximity, over-amplify geopolitical favouritism at inference time.

Long document retrieval aims to fetch query-relevant documents from a large-scale collection, where knowledge distillation has become de facto to improve a retriever by mimicking a heterogeneous yet powerful cross-encoder. However, in contrast to passages or sentences, retrieval on long documents suffers from the scope hypothesis that a long document may cover multiple topics. This maximizes their structure heterogeneity and poses a granular-mismatch issue, leading to an inferior distillation efficacy. In this work, we propose a new learning framework, fine-grained distillation (FGD), for long-document retrievers. While preserving the conventional dense retrieval paradigm, it first produces global-consistent representations crossing different fine granularity and then applies multi-granular aligned distillation merely during training. In experiments, we evaluate our framework on two long-document retrieval benchmarks, which show state-of-the-art performance.

Climate change has increased the intensity, frequency, and duration of extreme weather events and natural disasters across the world. While the increased data on natural disasters improves the scope of machine learning (ML) in this field, progress is relatively slow. One bottleneck is the lack of benchmark datasets that would allow ML researchers to quantify their progress against a standard metric. The objective of this short paper is to explore the state of benchmark datasets for ML tasks related to natural disasters, categorizing them according to the disaster management cycle. We compile a list of existing benchmark datasets introduced in the past five years. We propose a web platform - NADBenchmarks - where researchers can search for benchmark datasets for natural disasters, and we develop a preliminary version of such a platform using our compiled list. This paper is intended to aid researchers in finding benchmark datasets to train their ML models on, and provide general directions for topics where they can contribute new benchmark datasets.

How do we design measures of social bias that we trust? While prior work has introduced several measures, no measure has gained widespread trust: instead, mounting evidence argues we should distrust these measures. In this work, we design bias measures that warrant trust based on the cross-disciplinary theory of measurement modeling. To combat the frequently fuzzy treatment of social bias in NLP, we explicitly define social bias, grounded in principles drawn from social science research. We operationalize our definition by proposing a general bias measurement framework DivDist, which we use to instantiate 5 concrete bias measures. To validate our measures, we propose a rigorous testing protocol with 8 testing criteria (e.g. predictive validity: do measures predict biases in US employment?). Through our testing, we demonstrate considerable evidence to trust our measures, showing they overcome conceptual, technical, and empirical deficiencies present in prior measures.

Data Augmentation (DA) is frequently used to automatically provide additional training data without extra human annotation. However, data augmentation may introduce noisy data that impairs training. To guarantee the quality of augmented data, existing methods either assume no noise exists in the augmented data and adopt consistency training or use simple heuristics such as training loss and diversity constraints to filter out ``noisy'' data. However, those filtered examples may still contain useful information, and dropping them completely causes loss of supervision signals. In this paper, based on the assumption that the original dataset is cleaner than the augmented data, we propose an on-the-fly denoising technique for data augmentation that learns from soft augmented labels provided by an organic teacher model trained on the cleaner original data. A simple self-regularization module is applied to force the model prediction to be consistent across two distinct dropouts to further prevent overfitting on noisy labels. Our method can be applied to augmentation techniques in general and can consistently improve the performance on both text classification and question-answering tasks.

When working with the dataset, we noticed an imbalance across the classes, in the train-We apply the vision transformer, a deep machine learning model (Figure 1a) as well as in the provided devel-dataset. The counts build around the attention mechanism, on mel-spectrogram per sample and class are similar in train & devel (c.p. Figure 1a), representations of raw audio recordings. When adding melbased but we noticed slight variations in audio-sample length (number data augmentation techniques and sample-weighting, we of frames, c.p. Figure 1b & 1c). Distributions for train and devel achieve comparable performance on both (PRS and CCS challenge) are comparable while the test dataset has variations in sample tasks of ComParE21, outperforming most single model length regarding the number of very small (<= 0.3 seconds).

Opinion mining is the branch of computation that deals with opinions, appraisals, attitudes, and emotions of people and their different aspects. This field has attracted substantial research interest in recent years. Aspect-level (called aspect-based opinion mining) is often desired in practical applications as it provides detailed opinions or sentiments about different aspects of entities and entities themselves, which are usually required for action. Aspect extraction and entity extraction are thus two core tasks of aspect-based opinion mining. his paper has presented a framework of aspect-based opinion mining based on the concept of transfer learning. on real-world customer reviews available on the Amazon website. The model has yielded quite satisfactory results in its task of aspect-based opinion mining.

Evaluation is the central means for assessing, understanding, and communicating about NLP models. In this position paper, we argue evaluation should be more than that: it is a force for driving change, carrying a sociological and political character beyond its technical dimensions. As a force, evaluation's power arises from its adoption: under our view, evaluation succeeds when it achieves the desired change in the field. Further, by framing evaluation as a force, we consider how it competes with other forces. Under our analysis, we conjecture that the current trajectory of NLP suggests evaluation's power is waning, in spite of its potential for realizing more pluralistic ambitions in the field. We conclude by discussing the legitimacy of this power, who acquires this power and how it distributes. Ultimately, we hope the research community will more aggressively harness evaluation for change.